Yifan Wang

DRIFT: Learning from Abundant User Dissatisfaction in Real-World Preference Learning

LLMs Can Get "Brain Rot"!

Cascade Reward Sampling for Efficient Decoding-Time Alignment

More is Less: The Pitfalls of Multi-Model Synthetic Preference Data in DPO Safety Alignment

From Low Rank Gradient Subspace Stabilization to Low-Rank Weights: Observations, Theories, and Applications

Cascade Reward Sampling for Efficient Decoding-Time Alignment